From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: with ECARTIS (v1.0.0; list xfs); Mon, 27 Oct 2008 06:47:30 -0700 (PDT) Received: from cuda.sgi.com (cuda3.sgi.com [192.48.176.15]) by oss.sgi.com (8.12.11.20060308/8.12.11/SuSE Linux 0.7) with ESMTP id m9RDlH7u026136 for ; Mon, 27 Oct 2008 06:47:17 -0700 Received: from bombadil.infradead.org (localhost [127.0.0.1]) by cuda.sgi.com (Spam Firewall) with ESMTP id C9DE51488026 for ; Mon, 27 Oct 2008 06:47:16 -0700 (PDT) Received: from bombadil.infradead.org (bombadil.infradead.org [18.85.46.34]) by cuda.sgi.com with ESMTP id uo2c71pxUm5GC0ig for ; Mon, 27 Oct 2008 06:47:16 -0700 (PDT) Received: from hch by bombadil.infradead.org with local (Exim 4.68 #1 (Red Hat Linux)) id 1KuSRA-0001Yf-GK for xfs@oss.sgi.com; Mon, 27 Oct 2008 13:47:16 +0000 Date: Mon, 27 Oct 2008 09:47:16 -0400 From: Christoph Hellwig Subject: [PATCH 04/10] reduce l_icloglock roundtrips Message-ID: <20081027134716.GE5730@infradead.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline; filename=xfs-reduce-log-lock-roundtrips Sender: xfs-bounce@oss.sgi.com Errors-to: xfs-bounce@oss.sgi.com List-Id: xfs To: xfs@oss.sgi.com All but one caller of xlog_state_want_sync drop and re-acquire l_icloglock around the call to it, just so that xlog_state_want_sync can acquire and drop it. Move all lock operation out of l_icloglock and assert that the lock is held when it is called. Note that it would make sense to extende this scheme to xlog_state_release_iclog, but the locking in there is more complicated and we'd like to keep the atomic_dec_and_lock optmization for those callers not having l_icloglock yet. (First sent on Semptember 29th) Signed-off-by: Christoph Hellwig Index: linux-2.6-xfs/fs/xfs/xfs_log.c =================================================================== --- linux-2.6-xfs.orig/fs/xfs/xfs_log.c 2008-09-29 10:54:19.000000000 +0200 +++ linux-2.6-xfs/fs/xfs/xfs_log.c 2008-09-29 10:58:23.000000000 +0200 @@ -724,8 +724,8 @@ xfs_log_unmount_write(xfs_mount_t *mp) spin_lock(&log->l_icloglock); iclog = log->l_iclog; atomic_inc(&iclog->ic_refcnt); - spin_unlock(&log->l_icloglock); xlog_state_want_sync(log, iclog); + spin_unlock(&log->l_icloglock); error = xlog_state_release_iclog(log, iclog); spin_lock(&log->l_icloglock); @@ -762,9 +762,9 @@ xfs_log_unmount_write(xfs_mount_t *mp) spin_lock(&log->l_icloglock); iclog = log->l_iclog; atomic_inc(&iclog->ic_refcnt); - spin_unlock(&log->l_icloglock); xlog_state_want_sync(log, iclog); + spin_unlock(&log->l_icloglock); error = xlog_state_release_iclog(log, iclog); spin_lock(&log->l_icloglock); @@ -1957,7 +1957,9 @@ xlog_write(xfs_mount_t * mp, if (iclog->ic_size - log_offset <= sizeof(xlog_op_header_t)) { xlog_state_finish_copy(log, iclog, record_cnt, data_cnt); record_cnt = data_cnt = 0; + spin_lock(&log->l_icloglock); xlog_state_want_sync(log, iclog); + spin_unlock(&log->l_icloglock); if (commit_iclog) { ASSERT(flags & XLOG_COMMIT_TRANS); *commit_iclog = iclog; @@ -3166,7 +3168,7 @@ try_again: STATIC void xlog_state_want_sync(xlog_t *log, xlog_in_core_t *iclog) { - spin_lock(&log->l_icloglock); + ASSERT(spin_is_locked(&log->l_icloglock)); if (iclog->ic_state == XLOG_STATE_ACTIVE) { xlog_state_switch_iclogs(log, iclog, 0); @@ -3174,10 +3176,7 @@ xlog_state_want_sync(xlog_t *log, xlog_i ASSERT(iclog->ic_state & (XLOG_STATE_WANT_SYNC|XLOG_STATE_IOERROR)); } - - spin_unlock(&log->l_icloglock); -} /* xlog_state_want_sync */ - +} /***************************************************************************** --